Mass modeling of two apple varieties by geometrical attributes

In this study mass of two Iranian apple varieties (Golab Kohanz and Shafi Abadi) were predicted using different physical characteristics in linear models as three different classifications: (1) single or multiple variable regressions of apple dimensional characteristics. (2), single or multiple variable regressions of apple projected areas and (3), estimating apple mass based on its volume. All properties considered in the current study were found to be statistically significant at the 1% probability level. The performance of the models was evaluated by a set of test data using three different measures including the root mean square error (RMSE), the coefficient of determination (R2) and the mean error (ME) between predicted and measured values. In the first classification among single variable models the model based on width (M= 3.29^-116.25) had maximum coefficient of determination, R2=0.91 for Golab Kohanz variety and the model based on thickness (M=3.29T-118.06) had maximum coefficient of determination, R2=0.91 for Shafi Abadi variety, respectively. Among all of single variable models, the mass model based on actual volume was the best model for both varieties. For multiple variable regression models the best models were based on three dimension and three projected areas for all observation.